Publication Details
Cross-Domain and Cross-Language Portability of Acoustic Features Estimated by Multilayer Perceptrons
Grézl František, Ing., Ph.D. (DCGM FIT BUT)
Hwang Mei-Yuh (UWASH)
Lei Xin (UWASH)
Morgan Nelson, Prof. (ICSI Berkeley)
Vergyri Dimitra (UC Berkeley)
Cross domains, cross language, portability, probabilistic features, MLP features
Cross domains and language portability of phone-posterior features. English-trained MLP features can provide a significant
boost to recognition accuracy in new domains within the same
language, as well as in entirely different languages such as Mandarin and Arabic.
Recent results with phone-posterior acoustic features estimated by
multilayer perceptrons (MLPs) have shown that such features can
effectively improve the accuracy of state-of-the-art large vocabulary
speech recognition systems. MLP features are trained discriminatively
to perform phone classification and are therefore,
like acoustic models, tuned to a particular language and application
domain. In this paper we investigate how portable such features
are across domains and languages. We show that even without
retraining, English-trainedMLP features can provide a significant
boost to recognition accuracy in new domainswithin the same
language, as well as in entirely different languages such as Mandarin
and Arabic. We also show the effectiveness of feature-level
adaptation in porting MLP features to new domains.
@INPROCEEDINGS{FITPUB8248,
author = "Andreas Stolcke and Franti\v{s}ek Gr\'{e}zl and Mei-Yuh Hwang and Xin Lei and Nelson Morgan and Dimitra Vergyri",
title = "Cross-Domain and Cross-Language Portability of Acoustic Features Estimated by Multilayer Perceptrons",
pages = "321--324",
booktitle = "2006 IEEE International Conference on Acoustic, Speech, and Signal Processing",
year = 2006,
location = "Toulouse, FR",
publisher = "IEEE Signal Processing Society",
ISBN = "978-3-540-74627-0",
language = "english",
url = "https://www.fit.vut.cz/research/publication/8248"
}